Transforming mold line data ... into process knowledge: translating collected data on your nobake mold line into knowledge will aid your facility in identifying opportunities where improvements can be made and time and resources can be saved.
Durametal Corp., Muncy, Pa., a nobake ferrous casting firm, wanted to transform its mold line process data into useful information for plant operations. Knowledge about the mold line process would help identify opportunities where improvements can be made, process capability can be established, and time and resources can be saved.
Previously, Durametal's mold line data were distributed in various locations throughout the facility and computer network. Some data were kept digitally on computer files, but other data were kept on paper. Obtaining a true picture of the process operating conditions was difficult, and plant managers questioned the reliability of the information due to the discontinuous nature of the data collection.
The firm set a few project goals for transforming its mold line data into process knowledge. It wanted to establish an accessible and central location for mold line data, a system with a simple interface for data entry and retrieval and a method to display an overall look at current mold conditions through the collected data. Once these objectives were met, the foundation for further process analysis was in place to pursue its quality objectives.
Durametal's new mold line data application incorporated Microsoft Excel with Visual Basic for Applications (VBA) programming. Metalcasting operators and engineering personnel were granted access to the mold line data application on the facility's computer network to make data from the shop floor more widely available.
The mold line data application is comprised of two parts: data entry and data presentation. Both parts create opportunities for further data analysis using statistical tools, such as control charts. The data entry forms, which accommodate each set of data, were built in Excel and are used by the operator. Once an operator enters data onto a data entry form, they click a button to activate VBA code that transfers the data to a pre-designated spreadsheet for the particular data set.
The operators' computer interaction was made as streamlined as possible. An operator usually performs only a few mouse clicks to open a data entry form, and only a couple more mouse clicks are required to save data and exit the application.
Once data are actively being collected, Excel charts are set up to show a simple chronological display of selected parameter data.
Control charts are created from data collected within the mold line data application. The control charts, which serve as a method to track mold quality and generate information from data, consist of four parts: the centerline (the process mean), collected process data versus time and calculated upper and lower control lines. Upper and lower control limits (UCL and LCL) are limits to maximum expected variation of the process.
When the process is shown to be running within control limits, it is within statistical control. When the process goes outside control limits, this indicates the presence of assignable cause variation, which contributes the largest portion of overall variability in an unstable process.
Identifying this occurrence offers the opportunity to make improvements to process uniformity. (It should be noted, however, that a control chart indicating a process is in control does not necessarily assure the process will produce quality output.)
The mold line data application uses the following recorded data to create control charts from facing sand data:
* ratio of work time vs. strip time;
* grain fineness number (GFN);
* fines percentage;
* sieve screen distribution, numbers 70, 100 and 140;
* sand temperature.
Results: Transforming Data into Information
Quality mold sand was thought to be one of the leading factors affecting overall mold quality and scrap rate at Durametal, so determining sand quality was an objective of the mold line data analysis.
Likely candidates that could influence mold scrap include binder input levels, catalyst input levels and the work strip ratio. Work strip ratio data, which reflect the conditions of the mold sand, were chartered over a three-month span. During this time, work strip data was shown to be out of control in two different instances (Fig. 1). The out-of-control occurrence on Aug. 16 seems to be statistically insignificant, because data points before and after do not show any similar trend. But out-of-control data readings for Aug. 30 through Sept. 2 show a surge in work-strip ratio data during a seven-day period of operation.
[FIGURE 1 OMITTED]
An increase in work-strip ratio indicates that work time is increasing faster than strip time. A look into GFN data, fines data and sieve screen data could yield some insight into how sand quality affects work-strip time ratio data.
Investigating the Data
The GFN data set (Fig. 2) charted over the same three-month span as work-strip ratio data shows that average grain size remained in control and stable during the period when work-strip ratio data was trending out of control. GFN data alone indicate that the mold sand appears to be in good condition and not responsible for the out-of-control work-strip data.
[FIGURE 2 OMITTED]
Fines data, calculated from the total content from 200-screen size and smaller, actually were showing an overall decrease in content during the time period when work-strip data showed out-of-control data readings (Fig. 3). Seemingly, the courser sand would mean a shorter work time, which does not make sense. The fines data appear to be mostly inconclusive and likely not a large contributing factor for the short period that work-strip data were out of control.
[FIGURE 3 OMITTED]
Durametal's sand is considered "three-screen sand," with the largest percentage of sand separating out on 70, 100 and 140 screens. Sieve screen data give a general idea of the grain size distribution the system provides. Data recorded from three screens were used to construct control charts over the June through September three-month span.
The charts reveal that data from the 100-screen setup have a relatively small amount of variation, but trend lower for most of the data set before assuming a new average in the last month.
The charts also show that after the plant shut down for maintenance in late July, the data from the 70-screen setup became less variable and started to trend downward, similar to 100-screen data. Interestingly, the data from the 140-screen setup show the same amount of variability and are almost the inverse of 70-screen data throughout the entire dataset. It appears that for every drop in percentage on 140-screen data, the 70-screen data pick up the slack and shifts the sand volume, indicating a strong correlation between the two data sets.
Because 70 and 140 screens are on the ends of the distribution scale for the three chief sand screens, when the two values oscillate, mold sand tends to move between a sand content that is either coarse or fine. This knowledge about the variability of the process may not have been apparent without the data management system.
Still, sieve screen data show a trend that becomes coarser, which agrees with the fines data, but shows no correlation to work-strip. During the out-of-control work-strip data event, screen data remain within statistical control and possess no assignable cause variation.
Another factor related to sand condition is sand temperature. Because binder catalyst is exponentially sensitive to temperature changes, a control chart was made from sand temperature data collected from the mold line (Fig. 4). Sand temperature appears to be one of the most volatile factors surrounding the process. An investigation into better sand temperature may lead to a more stable process.
[FIGURE 4 OMITTED]
The sand temperature control chart shows that during the period when work-strip data were "out of control," the sand was close to the mean temperature of 90.9 degrees and does not show a recognizable trend that would indicate the presence of assignable cause variation.
Because the analysis of mold line application data pertaining to work-strip ratio out-of-control data points cannot be correlated to a lack of sand quality with data currently being compiled, sand quality may not be the problem. Other process inputs, such as mold sand binder, smart pumps that regulate binder input, or simply measurement error, could be the cause of the out-of-control work-strip event.
The possibility also exists that the new data collection method is not good enough to capture all factors that indicate sand quality. Perhaps something else should be monitored, such as the sand's pH balance and acid demand value. In addition, an investigation into mold scrap and its correlation to the work-strip ratio may help indicate how the sand quality is related to mold scrap.
Transforming Information into Knowledge
Future additions to the mold line application might include recording more data relating to patterns used to create molds. This will be one step toward extracting more knowledge from the data. Tracking which patterns are scrapped and why, in addition to tracking the mold line parameter levels at the point a mold is scrapped, may aid in determining why certain patterns are more susceptible to being scrapped. This could lead to information that yields the optimum settings molds require for molding success.
A system could be established that prescribes mold line parameter settings to optimize mold making for the particular patterns. At Durametal, demand is widely spread over a large number of patterns used to produce molds daily. Pattern ratings in conjunction with casting demand could dictate a molding schedule that acts as a plan for mold line production so the mold line process can:
* accommodate all varied geometries of patterns demanded;
* help to eliminate mold scrap;
* pass quality on to the following stages of production.
This material is based upon work supported by the Department of Energy (DOE) under Award No. DE-FC07-02ID14228. Any opinions, findings or conclusions and recommendations expressed in this article are those of the authors and do not necessarily reflect the views of the DOE.
This article was adapted from a paper at the Steel Founders' Society of America 2005 Technical and Operating Conference, November 2005, Chicago.
About the Authors
Greg Saveraid is an undergraduate student and Frank Peters is an associate professor in the Industrial and Manufacturing Systems Engineering Dept. at Iowa State Univ., Ames Iowa. John Cory is a process engineer for Durametal Corp., Muney, Pa.
For More Information
"Maximizing Productivity of Nobake Molding Operations," T. J. Gilbreath, MODERN CASTING, Aug. 2000, p. 39-42.
|Printer friendly Cite/link Email Feedback|
|Date:||Jan 1, 2006|
|Previous Article:||Consolidating your cleaning room: by implementing new designs and practices, two steel casters anticipate improvement in their cleaning and finishing...|
|Next Article:||How do manganese, sulfur levels affect gray iron properties?|